Analytical Approaches To World Music 2.1 (2012) 87-137 CantoCore: A New Cross-Cultural Song Classification Scheme

Patrick E. Savage, Emily Merritt, Tom Rzeszutek, and Steven Brown Department of Psychology, Neuroscience & Behaviour, McMaster University

Classification of organisms and languages has long provided the foundation for studying biological and cultural history, but there is still no accepted scheme for classifying songs cross-culturally. The best candidate, Lomax and Grauer’s “Cantometrics” coding scheme, did not spawn a large following due, in part, to concerns about its reliability. We present here a new classification scheme, called “CantoCore”, that is inspired by Cantometrics but that emphasizes its “core” structural characters rather than the more subjective characters of performance style. Using both schemes to classify the 30 songs from the Cantometrics Consensus Tape, we found that CantoCore appeared to be approximately 80% more reliable than Cantometrics. Nevertheless, Cantometrics still demonstrated significant reliability for all but its instrumental characters. Future multidisciplinary applications of CantoCore and Cantometrics to the cross-cultural study of musical similarity, musical evolution, musical universals, and the relationship between music and culture will provide the true test of each scheme’s value.

Keywords: music, song, classification, CantoCore, Cantometrics, reliability, musilinguistic continuum

This is a single-spaced version of the article. The official version with page numbers 87-137 can be viewed at http://aawmjournal.com/articles/2012a/Savage_Merritt_Rzeszutek_Brown_AAWM_Vol_2_1.pdf

INTRODUCTION according to geographic and ethnolinguistic usical classification is a topic that classifications do not use an explicitly musical has received scant attention since classificatory framework. the heyday of comparative A consideration of the historical roots of the musicology during the first half of field shows that classification was central to the first the 20th century. Fields like definition of comparative musicology: biologyM and linguistics have long relied on [C]omparative musicology has as its task the classification as the starting point for developing comparison of the musical works— broader theories, such as Darwin’s (1859) theory of especially the folksongs—of the various evolution and Jones’ (1807) theory of prehistoric peoples of the earth for ethnographical connections among speakers of Indo-European purposes, and the classification of them languages. Today, global linguistic classification according to their various forms (Adler databases such as the Ethnologue (Lewis 2009) and 1885, 14). the World Atlas of Language Structures Although classification, comparison, and (Haspelmath et al. 2005) are fundamental to the ethnography were all equal parts of this original study of language evolution, linguistic universals, definition, the field later changed its name to and human history (Currie and Mace 2009; Dunn et “ethnomusicology” and developed a methodological al. 2011; Atkinson 2011). Musicology, in contrast, emphasis on single-culture ethnography over cross- never entered into a comfortable relationship with cultural classification and comparison. This was cross-cultural classification, despite early attempts part of a broader trend in anthropology in the wake in that direction (Hornbostel and Sachs 1914; of World War II toward cultural relativism and Lomax 1968). Even global music collections like away from universalism (Geertz 1973). One the Garland Encyclopedia of World Music (Stone et outcome of this shift was the recognition of a al. 1998) and Smithsonian Global Sound theoretical distinction between “etic” (objective, (http://glmu.alexanderstreet.com) that are organized outsider) and “emic” (subjective, insider) theories of 1

CantoCore: A New Cross-Cultural Song Classification Scheme classification (Harris 1976). This dichotomy nicely based on multidimensional, musilinguistic spectra characterizes the paradigmatic difference between could be helpful in fields as diverse as early comparative musicology and contemporary ethnomusicology, neuroscience, and evolutionary ethnomusicology. Ethnomusicologists have largely biology for understanding connections between rejected etic and/or acoustic classification schemes, music and language (Darwin 1871; Feld and Fox despite pleas for pluralism in approaches to world 1994; Wallin, Merker, and Brown 2000; Patel musics (Merriam 1982; Nettl 2005; Agawu 2010). 2008). Although the goal of classifying musics acoustically Multi-dimensional, musilinguistic spectra are in presents many challenges—for example, the need fact a major design feature of the best-established that classification schemes be universally song-classification scheme to date, “Cantometrics” applicable—these challenges do not a priori (Lomax and Grauer 1968; Lomax 1976). invalidate cross-cultural classification (but see Hood Cantometrics classifies songs according to 37 1971; Blacking 1973; McLeod 1974). acoustic characters related to their structure, Along these lines, there are two major performance style, and instrumental methodological challenges to classifying music accompaniment. Each character contains between 3 cross-culturally. One challenge is specific to and 13 character-states, which are ordered along a instrumental music: how do we ensure that we are social continuum from “individualized” to comparing like with like when different cultures use “groupy.” This continuum can be thought of equally different instruments with differing acoustic well as a musilinguistic continuum, since speech features, production mechanisms, and tuning tends to be more individual-oriented and song more systems (Ellis 1885)? The second is specific to group-oriented. vocal music: how can we design a classification Applying this scheme to a global sample of system that is broad enough to accommodate all thousands of songs from hundreds of cultures, musical cultures while maintaining a distinction Lomax found that global song diversity was between “song” and “speech”? While the organized into 10 major stylistic families that also instrumental classification scheme of Hornbostel correlated with extra-musical features of social and Sachs (1914) is still widely used today, there structure and historical contact. Critics generally remains no widely accepted song-classification applauded this ground-breaking attempt to scheme. quantitatively address the relationship between One solution to the problem of song music and culture and supported its broad findings, classification is to see the relationship between despite some concerns over methodological issues music and language as a continuum—a regarding sampling, treatment of intra-cultural “musilinguistic” spectrum (Brown 2000)—rather diversity, and the interpretation of correlations than as a contrast between two discrete domains. A between music and social structure (Naroll 1969; truly universal approach cannot exclude “non- Driver 1970; Downey 1970; Nettl 1970; Maranda musical” vocalizations but must accommodate any 1970; Henry 1976; Erickson 1976; Dowling and type of vocalization sitting along the musilinguistic Harwood 1986; Grauer 2005; Leroi and Swire spectrum of communicative forms from speech, to 2006). However, many critics were divided over songs, to everything in between. While Sachs Lomax’s emphasis on performance style over song (1943) proposed such a spectrum in his distinction structure. Lomax’s agenda in creating Cantometrics between “logogenic” (word-born) and “melogenic” was to replace Western musicology’s traditional (melody-born) songs, there is a need for a emphasis on musical structure and notation—which classification scheme that can accommodate the he and many others saw as being Eurocentric and diversity of ways in which song-features can elitist (Lomax 1959; Feld and Fox 1994)—with a independently vary across multiple musilinguistic more performance-oriented system. While some spectra. For example, some songs can have irregular critics supported the development of measurements “speech-like” (parlando) rhythms but use discrete of performance characters such as “nasality” and “music-like” pitches, while others can have metric “rasp,” others were concerned that such characters “music-like” rhythms but use indeterminate were overly subjective and thus unreliable (Downey “speech-like” pitches. A classificatory approach 1970; Maranda 1970). 2 Analytical Approaches To World Music 2.1 (2012) 87-137 The principal objective of the current study is to structural characters not present in Cantometrics, present a detailed analysis of a new universal song- most notably those related to scales and rhythms. classification scheme. We call it “CantoCore” Finally, the scheme is designed to accommodate because of its emphasis on the “core” structural musical forms at all points along the musilinguistic characters of song. The scheme takes its lead from spectrum, from a simple sentence to the most the updated 1976 version of Cantometrics but complexly-textured responsorial polyphony. The focuses only on characters of song-structure rather current study also includes a test of the inter-rater than performance-style or instrumentation (see reliability of song codings, comparing 1) CantoCore Figure 1), because of our prediction that structural vs. Cantometrics, and 2) the structural characters of characters should be more reliable. We have Cantometrics vs. its performance and instrumental reorganized, supplemented, and attempted to more characters. To accomplish this, we use the global set objectively operationalize these characters, building of 30 songs contained in the Cantometrics on the work of others whenever possible (Kolinski Consensus Tape (Lomax 1976) that Lomax 1961, 1962, 1973; Plomp and Levelt 1965; Patel selected to demonstrate the cross-cultural validity of and Daniele 2003; Leroi and Swire 2006; Busby the Cantometrics scheme. 2006). In addition, the scheme introduces several

Figure 1. A comparison of the types of musical characters classified by CantoCore vs. Cantometrics. Both classification schemes rely exclusively on acoustic information rather than on non-acoustic characters. Whereas Cantometrics (green box) focuses on both the performance and structural characters of songs as well as their instrumental accompaniment, CantoCore (red box) focuses exclusively on the structural characters of the vocal part, excluding both performance and instrumental characters.

3

CantoCore: A New Cross-Cultural Song Classification Scheme hierarchy that we employ in organizing the CLASSIFICATION SCHEME characters of the CantoCore classification scheme; the characters themselves are listed in Figure 2b. Theoretical Framework A useful analogy for conceptualizing our The musical hierarchy. Music is a hierarchical classification scheme is to think of a song as a system made up of several levels of organization biological organism. In essence, songs are simply (Schenker 1979; Lerdahl and Jackendoff 1983; complex combinations of notes, just as organisms Krumhansl 1990; Anku 2000; Tenzer 2006). Figure are complex combinations of cells. However, as 2a presents a schematization of the musical with the cells in an organism, the notes in a song a) Phrase#1 Phrase#2 Note#domain: 3 Rhythm Part#1 3 Pitch la 3 Syllable

la la la la Supra3note domain: 3note) Part#2 3part) 3phrase) Note#1 Note#2 b) 1)#Meter 2)#Number of#beats 3)#Beat#sub3division Rhythm 4)#Number#of#sub3beats 5)#Syncopation 6)#Motivic redundancy 7)#Durational#variability 8)#Tonality 9)#Mode 10)#Number#of#pitch#classes (between3note) Pitch 11)#Hemitonicity 12)#Melodic#interval#size 13)#Melodic range 14)#Melodic#contour 15)#Melisma Syllable 16)#Vocables 17)#Number#of#vocal#parts 18)#Rhythmic#texture 19)#Harmonic#texture (between3part) 20)#Relative#motion 21) Phrase#repetition 22)#Phrase#length 23)#Phrase#symmetry (between3phrase) 24)#Solo/group#arrangement 25)#Responsorial#arrangement 26)#Phrase overlap Figure 2. a) The musical hierarchy is comprised of “note” and “supra-note” domains. The three main note domains are rhythm (red), pitch (blue), and syllable (green), as represented by the sung note “la.” Interactions between notes give rise to the supra-note domains of “phrase” (the between-note level), “texture” (the between-part level) and “form” (the between-phrase level). b) The 26 structural characters that comprise the CantoCore classification scheme are organized4 according to these note and supra-note domains.

Analytical Approaches To World Music 2.1 (2012) 87-137 interact with each other and with their extra-musical ranging from being completely absent (low environment at many different levels and in many frequency) to being ubiquitous (high frequency). different ways. These complex interactions can Qualitative traits, by contrast, cannot be placed onto never be fully quantified but can still be usefully a numerical spectrum of size or frequency, and are modeled. instead organized as a series of discrete states. For The most basic distinction is that between the example, melodic contours (character 14) come in a note level—where the note is regarded as the basic variety of discrete types, such as descending building block of music—and the supra-note level. contours, ascending contours, arched contours, and The note level consists of three characters: 1) the like. Of the 26 CantoCore characters, 15 are rhythm (colored red in Figure 2a), reflecting the quantitative traits and 11 are qualitative traits by the relative duration of a note; 2) pitch (blue), reflecting standards of classification theory. the acoustic frequency of a note; and 3) syllable Ordering of character-states. Most of (green), reflecting the articulatory configuration of a CantoCore’s 26 characters are divided into 3–4 sung note (exemplified by “la” in the figure). The character-states, resulting in a total of 96 character- supra-note domain consists of interactions between states across the scheme. Of these, 53 are new to the notes, as organized into three broad hierarchical scheme, as indicated by asterisks in the detailed domains: 1) phrase, representing the between-note Figure 3 represents our rationale for ordering the level within individual vocal parts, 2) texture, character-states within each character. Character- representing the between-part level, in which states are ordered in a consistent manner, spanning a simultaneous phrases in different vocal parts musilinguistic spectrum from language-like (left overlap in time, and 3) form, representing the side) to music-like (right side). However, the between-phrase level, where successive phrases method for achieving this differs for quantitative combine to form larger melodic units. Figure 2b and qualitative characters, as shown in Figure 3 lists the classification characters associated with above and below the horizontal arrow. For each of these three supra-note domains. It also quantitative characters, character-states are listed in shows that the domain of “phrase” contains the order of increasing size or frequency using lower- three note-level characters of rhythm, pitch, and case roman numerals (i, ii, iii, etc). This allows for syllable (color coded the same as in Figure 2a). precise placement of states along a continuum CantoCore classifies 26 structural characters of spanning from small (speech-like) to large (song- songs (Figure 2b), organized into categories like). For qualitative characters, character-states are associated with the note and supra-note domains listed in order of increasing “regularity” using listed above. Fifteen of these characters are refined lower-case letters (a, b, c, etc.), spanning from versions of structural characters already contained irregular (speech-like) to regular (song-like). By in Cantometrics, while 11 characters—mostly those regularity, we refer to the degree of repetitiveness of related to rhythm and scale—are new, as indicated a character throughout a song, where redundancy is by asterisks in the detailed scheme below. far more associated with music than speech (Lomax Quantitative vs. qualitative characters. A 1968). Because qualitative characters could not fundamental distinction in classification theory is always be divided up consistently, we employed a that between quantitative (or continuous) characters series of prefixes to convey a spectrum of and qualitative (or discrete) characters (Sneath and qualitative states (see the geometric shapes at the Sokal 1973). Quantitative traits can be classified bottom half of Figure 3 as a guide): a) “A-” implies with regard to their size. For example, melodic that a feature is absent from a song; b) “Hetero-” intervals (character 12 in the CantoCore scheme) implies that multiple but successive features occur; vary in a continuous manner from very small c) “Poly-” implies that multiple simultaneous intervals to very large, and everything in between. features occur; and d) “Iso-” implies that a single Another way to code characters quantitatively is feature occurs consistently throughout a song. with regard to their frequency of occurrence in a Applying these concepts to meter, for example, we song. In CantoCore, vocables (character 16) are can see that irregular “a-metric” songs have no coded with regard to their frequency of occurrence, discernable meter; semi-regular “hetero-metric” or 5

CantoCore: A New Cross-Cultural Song Classification Scheme

Figure 3. The character-states within each character are organized according to a “musilinguistic” spectrum spanning from language- like to music-like (no value judgment is implied). Quantitative characters (top part of the figure) are ordered in terms of increasing size from small to large using lower-case roman numerals. Qualitative characters (bottom part of the figure) are ordered in terms of increasing “regularity” using lower-case letters from irregular (“A-”) to regular (“Iso-”), with semi-regular states between them having either multiple successive forms (“Hetero-”) or multiple simultaneous forms (“Poly-”). The geometric shapes are used for heuristic purposes only to demonstrate the various facets of regularity.

“poly-metric” songs have multiple meters that are perception among Western and Javanese musicians, present successively or simultaneously, Perlman and Krumhansl (1996) found great respectively; and regular “iso-metric” songs have a variability within both groups of musicians. Even single, constant meter throughout. their most accurate subjects were limited by basic perceptual constraints in their ability to reliably Classification Logistics distinguish intervals differing by only 20 cents. On Classification by ear. The goal of our the other hand, one Javanese musician displayed classification system is to provide a tool to describe “regions of confusion” as large as 180 cents in and compare songs from many cultures in terms of which they perceived intervals from 120–300 cents multiple musical features. Ideally, one would want as being equivalent, presumably because they were to use an automatic acoustic-based classification using the slendro scale (which contains only one system or a database of musical scale degree in this range) as an internal interval transcriptions/notations to allow one to quickly and standard. objectively classify songs with a high degree of The greater the effects of each type of accuracy. Unfortunately, the automatic constraint, the lower the accuracy of classification classification systems and databases that currently by ear will be. Nevertheless, by dividing exist are heavily biased towards Western songs and quantitative characters into only three character- Western theory (e.g. Schaffrath 1995; Bertin- states, rather than the five character-states preferred Mahieux, Ellis, Whitman and Lamere 2011). Thus, in Cantometrics, we have tried to minimize the as with the creators of Cantometrics, we have been number of grey areas where such classificatory forced to develop a relatively blunt method that can ambiguities could occur while at the same time allow a coder to classify an individual song by ear maintaining the sense of a continuum of musical across several dozen features in a short amount of features, rather than a “presence/absence” time. Although we have tried to provide specific dichotomy. definitions and precise threshold values for all our The choice of precise threshold values is character states, these ultimately function as rough necessarily arbitrary, especially since there are no guidelines to help the coder reach a more holistic, comprehensive datasets other than Cantometrics subjective decision regarding the appropriate regarding the worldwide distribution of these classification. features. Therefore, we have tried to specify values The reliability of such classifications by ear is that will best capture the range of variation found limited both by lower-level perceptual constraints throughout the world, relying mainly on and by higher-level cognitive constraints. For Cantometrics and on our own subjective listening example, in an experiment testing interval experiences with world musics. For example, the 6 Analytical Approaches To World Music 2.1 (2012) 87-137 use of the perfect fifth and octave as thresholds for can only be accomplished by “multi-coding,” in “melodic range” (character 13) was maintained other words selecting multiple distinct character- from Cantometrics, while our choice of three and states for the same song (e.g., both “descending” five as thresholds for the number of pitch classes in and “arched” contours if both types occur in a single a scale (character 10) was based on our intuitions song). As a general rule, multi-coding should be that these would capture the most variation in scales avoided if one character-state is clearly the most throughout the world. prominent in a song. Within-song heterogeneity. Reality is too Character dependence. Some characters are complex to be fully captured in a single dependent on others. For example, “a-metric” songs classification. Songs change over time and can that have no beat (character 1) cannot possibly have contain multiple sections whose codings conflict a sub-beat (character 3). For such characters, an with one another. Some important work has been “n/a” character-state is included to denote done regarding quantifying this kind of dynamic something that is unclassifiable. A “?” may be used heterogeneity with regards to specific characters instead if recording quality or other factors make it such as interval size and note duration (Toiviainen impossible to code a given character, or if the and Eerola 2001; Huron 2006). However, there is musical characters are simply too complex to also a need for broader classification schemes that specify (following Busby 2006). provide simpler classifications but that span a Relationship to Cantometrics. For all characters number of characters across multiple domains. that are derived from structural characters of Maximal values. Heterogeneity can be partially Cantometrics, the original Cantometrics line accommodated for quantitative characters by number and corresponding character-states names defining them with regard to summary statistics from the updated version of Cantometrics (Lomax describing their size or frequency. Hence, a song 1976) have been given. There are a few small that has multiple states for such characters could be differences between this and the version used to coded with regard to things like their maximal value collect the original Cantometric data (Lomax and for that song, their mean value for the song, or their Grauer 1968), but these can be easily inter- standard deviation. For consistency, and to make the converted. Therefore, it is basically possible to scheme possible to use quickly by ear without convert old Cantometric codings into CantoCore resorting to laborious transcription and note- codings if desired, which may be useful in re- counting, quantitative characters have been defined analyzing the original Cantometric data without in terms of maximal values and divided into the having to re-code each of its thousands of songs. character-states of “small,” “medium,” and “large” Instrumental application. Due to the by imposing somewhat arbitrary thresholds. This is complications listed in the introduction involved in intended to reduce the amount of theoretical classifying instrumental music cross-culturally, we expertise and time required to code the songs. If one have designed CantoCore exclusively for the is working from notated scores or transcriptions, or purpose of classifying vocal music. Most of the if the coder has enough confidence in his/her ability classifications could also be useful for classifying to hear very fine distinctions, the raw numerical instrumental music, but caution should be exercised values may be used to increase precision (see Figure in doing so, particularly regarding the additional 4). However, this may give an appearance of constraints on sound production and intonation that precision that is unrealistic, as we found that are introduced by different instrument types. For making the scheme finer-grained did not improve its example, although breathing can still be helpful in reliability. determining phrase boundaries for aerophones, it Multi-coding. For qualitative characters, will be less useful when dealing with chordophones. heterogeneity is more difficult to classify. In some How to code. We have attempted to define all of cases, the heterogeneity of a song’s characters can our terms as precisely as possible so that the coder be accommodated by character-states that specify can provide precise numeric values if they are an intrinsic heterogeneity of features (e.g., “hetero- working directly from a score or transcription, or if metric,” “poly-tonal”). However, in other cases, this they have a high level of listening expertise. These 7

CantoCore: A New Cross-Cultural Song Classification Scheme definitions therefore require a modest background classify their own music). CantoCore is in music theory. However, since much of the fundamentally an etic, acoustic classification world’s music is transmitted orally and is difficult scheme, with all of the benefits and drawbacks that and time-consuming to transcribe, we have also this entails (Harris 1976). aimed to create our character-states so that they can Definitions. Our goal was to create a descriptive be reliably identified by ear without detailed system that allows a common vocabulary for notation. Ultimately, the numeric values are simply classification, not a prescriptive system that dictates guidelines to assist the coder in interpreting their how one should perceive music. Nevertheless, for holistic, subjective classification of the songs. Once such a system to be reliable, it is necessary to have the coder has practiced with a few dozen songs, standardized definitions. Since few, if any, musical he/she should be able to code a 3-minute song by terms have cross-culturally agreed-upon definitions, ear in 15–20 minutes, which is comparable to the we have offered our own definitions for each amount of time required to do so using character, as well as for several key terms (see Box Cantometrics (Lomax 1976). 1). Definitions about complex musical categories When coding, the coder should first listen to the such as “tonality” and even seemingly simpler song once through, jotting down important notes categories such as “interval size” have been, and and trying to get a sense for the different phrases will continue to be, debated. Our definitions are that make up the song: how many there are, in what simply operational ones that can be usefully applied order, how long each phrase is, what scale(s) or cross-culturally. Even when using these definitions, meter(s) (if any) underlie them, etc. The some level of disagreement and ambiguity is instrumental accompaniment can be used if it is inevitable due to perceptual differences between helpful in interpreting the correct song individuals and between cultures. We discuss some classification, but if there is any conflict between observations on agreement in the “Reliability” the vocal and the instrumental components, the section. coder should focus only on the vocal component. After they have listened to the song once, they THE “CANTOCORE” SONG CLASSIFICATION should go through and attempt to classify each SCHEME character in order from 1 to 26. They should then listen to the entire song again, checking the initial NOTE: Characters and character-states marked codings and paying particular attention to with an asterisk are those that are new to this complicated or ambiguous codings. The coder can scheme and that are not taken from Cantometrics. also jump forwards or backwards within the song or Modifications to original Cantometrics character- repeat the song as many times as necessary to arrive states are listed using parentheses. at a set of codings they are confident in. Any particularly noteworthy features, such as I) “PHRASE” (between-note) ambiguities or striking characters not classifiable, should be listed in a separate “comments” column. A) Rhythm This same format applies regardless of the length of the song or any extra-acoustic information 1) METER (Cantometrics Line 11) about the song. The definition of what constitutes a Cyclic, hierarchical groupings of beats into bars “song” varies, but in the absence of other (a) A-metric: No consistent beat (formerly information, it is reasonable to assume that different “parlando rubato – free rhythm”) tracks on recordings correspond to different songs. (b) Hetero-metric: There is a consistent beat, but Song classifications should be interpreted with the there is no consistent hierarchical pattern help of recording liner notes, music theory (both among these beats (formerly divided into “one- emic and etic), and all other available resources. beat rhythm” and “irregular meter”) However, the initial classification should be done (c) Poly-metric*: Multiple independent beats occur blind to extra-acoustic information as much as is simultaneously (e.g., 6/8 against 3/4, multiple practically possible (i.e., without knowing what singers singing in different tempi) (“simple” culture the song is from or how the singer(s) 8 Analytical Approaches To World Music 2.1 (2012) 87-137 and “complex” poly-meter from Cantometrics and the other of which is divided into three sub- Line 12 have been combined and moved here) beats. (d) Iso-metric: There is a single, consistent pattern of strong and weak beats (e.g., 3/4, 6/8, 5/4, Box 1: Glossary of key terms 2+2+3/8) (formerly divided into “simple” and “complex”) Note: A continuous combination of one pitch and one syllable for a fixed duration. If the pitch or syllable changes or begins N.B. See Box 1 for the definition of “beat.” afresh, this constitutes a new note. Songs not classified as (d) (“iso-metric”) must be coded (n/a) for characters (2–5). Songs that Vocal part: A series of notes sung by one voice, or by several transition between metric types (e.g., an “a-metric” voices in unison and/or in octaves. Slight variations between voices singing basically in unison are not counted as separate section giving way to an “iso-metric” section) parts unless the offset between parts exceeds 0.1s in time or should be multi-coded. 50 cents in pitch (see characters 18 and 19 in the scheme). Comments: The “poly-metric” character-state was moved here from Cantometrics Line 12. Phrase: A self-contained series of notes in one or multiple vocal parts. Phrases are usually separated by breaths or long Although it is debatable whether one can hear pauses, but can also be separated by more complex grouping multiple meters simultaneously (Kolinski 1973; principles. The coder should rely on their intuition in deciding London 2004), it is possible for a listener to what constitutes a new phrase, focusing on breaths in ambiguous cases. recognize the presence of two simultaneous meters/tempi and choose to attend to one or the Beat: Fixed time interval(s) at which notes regularly recur. other meter. Therefore, we have maintained this The beat is often sub-divided into multiple sub-beats. In cases character-state, although it may not be useful in the where the distinction between a “beat” and a “sub-beat” is ambiguous, the coder should designate the beat as the unit majority of cases. The new characters (2–4) were that feels the most natural to take steps to when dancing. created to deal with various iso-metric sub-types unclassifiable using Cantometrics. For instance, Tonic: The central tone(s) that seems to be the most stable in Cantometrics did not create any distinctions a scale. The tonic is usually either the most common note in a scale, the final note in a phrase, or both. In ambiguous cases, between 3/4, 4/4, 9/8 and 12/8 meters, although the coder should designate the tonic as the note that occurs there are important regional differences in the most frequently as the final note in a phrase. If the tonic distribution of these metric types. For example, 3/4 seems to consistently differ between phrases or between vocal and 9/8 rhythms are more common in Europe than parts, this should be classified as hetero- or poly-tonal, respectively (see character 8 in the scheme). in Africa or Asia, where 12/8 and 4/4 meters, respectively, are relatively more common (Stone et Pitch class: Notes that share the same note name (e.g., B, Db) al. 1998). regardless of their absolute pitch are considered as the same pitch class (i.e., assuming octave equivalence). Because the production of vocal pitches often fluctuates by up to 100 cents 2) NUMBER OF BEATS* from tonal targets during normal singing (Pfordresher et al. The number of beats in a bar 2010), we have followed the compromise adopted by Kolinski (a) Duple: The number of beats can be divided by (1961) and others of rounding pitches to the nearest 100 cents, for a maximum of 12 possible unique pitch classes. 2 (e.g., 2/4, 4/4, 6/8, 12/8, 2+3/8) Unfortunately, as a result of this compromise, there may be (b) Triple: The number of beats can be divided by some cases in which separate microtones are classified as a 3 but not by 2 (e.g., 3/4, 9/8, 2+2+3/8) single pitch class, while in other cases normal variation of (c) Complex: The number of beats can only be intonation may be classified as separate pitch classes.

divided by prime numbers greater than 3 (e.g., N.B. None of these terms have a well-agreed upon cross- 7/4, 5/8, 2+2+3+2+3/8) cultural definition. We offer these definitions to assist in (n/a) A-/hetero-/poly-metric: See (1) developing a shared classification vocabulary that can be Comments: Only the number of beats is coded reliably replicated by different coders. However, we recognize that many cultures have their own emic definitions here, regardless of the manner in which they are that may differ from ours, and that there are many grey areas sub-divided into sub-beats, which is coded in (3). in which the perception and interpretation of these features For example, a 2+3/8 meter is composed of two may vary both within and between cultures. beats, one of which is divided into two sub-beats

9

CantoCore: A New Cross-Cultural Song Classification Scheme 3) BEAT SUB-DIVISION* 5) SYNCOPATION* Division of beats into sub-beat-level metric The percentage of notes that are relatively groupings prominent (loud) but in metrically unaccented (a) A-divisive: Beats are not sub-divided (e.g., a positions 4/4 piece containing only q and h notes) (i) Little or no syncopation: <5% (b) Hetero-divisive: Beats are sub-divided, but the (ii) Moderately syncopated: 5–20% number of sub-beats per beat changes (e.g., (iii) Highly syncopated: >20% 2+2+3/8) (n/a) A-/hetero-/poly-metric: See (1) (c) Iso-divisive: Beats are sub-divided into a Comments: The term “syncopation” is used consistent number of sub-beats (e.g., 6/8, a 4/4 here instead of Kolinski’s (1973) term piece containing e notes) “contrametricity” because it is more widely (n/a) A-/hetero-/poly-metric: See (1) understood, because it allows us to recognize a N.B. See Box 1 for the distinction between continuum of varying degrees of syncopation rather “beat” and “sub-beat.” Songs not classified as (c) than a “commetric/contrametric” dichotomy, and (“iso-divisive”) must be coded (n/a) for character because Kolinski did not offer a precise definition (4). of contrametricity. Comments: This character was created to capture a crucial metric dimension not classified in 6) MOTIVIC REDUNDANCY* Cantometrics. It is almost identical to (2) but The percentage of all notes that are constructed captures a finer level of the metrical hierarchy and from a single recurring rhythmic pattern does not have a “poly-divisive” character-state (i) Non-motivic: <20% because this would be redundant with “poly-metric” (ii) Moderately motivic: 20–50% (see 1). (iii) Highly motivic: >50% N.B. If there are multiple motives, classify 4) NUMBER OF SUB-BEATS* based on the frequency of the most common motive. The number of sub-beats in a beat Comments: Figure 4 provides an example (a) Simple: The number of sub-beats can be where 40 out of the 61 notes (66%) are constructed

divided by 2 (e.g., q beat divided into e note from the rhythmic pattern e r r e sub-beats; includes 3/4, 4/4, etc.) (b) Compound: The number of sub-beats can be 7) DURATIONAL VARIABILITY* divided by 3 but not by 2 (e.g., q.k beat divided Maximum number of different types of into e note sub-beats; includes 6/8, 9/8) duration values in a song (c) Complex: The number of sub-beats can only be (i) Low durational variability: <3 duration values divided by prime numbers greater than 3 (e.g., q beat divided into 5 sub-beats) (e.g., only r and e) (n/a) A-/hetero-/poly-metric or a-/hetero-/poly- (ii) Moderate durational variability: 3–4 duration divisive: See (1/3) Comments: Songs in which groupings of five values (e.g., r, e and w) or more sub-beats are broken down into smaller (iii) High durational variability: >4 duration values groupings of twos and threes (e.g., 2+2+3+2+2/8 (e.g., , , , and ) [London 1995]) should be classified as “hetero- t r e w q divisive” songs (see 3) rather than “complex.” In N.B. Duration values refer to inter-onset “swing” time, sub-divisions that approximate 2:1 intervals (IOIs) as opposed to sounding durations should be classified as “compound,” while those (i.e., a quarter followed by an eighth rest and a that approximate 3:2 should be classified as dotted quarter both have the same IOI duration “complex.” value). Dotted notes are counted as separate duration values. Comments: Patel and Daniele (2003) present a different conception of rhythmic variability that focuses on variability between successive pairs of 10 Analytical Approaches To World Music 2.1 (2012) 87-137 notes, such that a series of all quarter notes has approach it (Powers et al. 2012). Nevertheless, we minimal variability and a series of alternating have chosen to focus on the major/minor distinction quarter notes and eighth notes has high variability. because it is commonly employed and relatively Our definition instead focuses on global variability amenable to classification. Characters dealing with across all notes in a song, which is easier to estimate micro-tonal intonations have been avoided due to a by ear and which provides a better means of lack of consensus on how to classify these examining differences between music and language, characters. whereas Patel & Daniele were explicitly trying to examine similarities between music and language. 10) NUMBER OF PITCH CLASSES* Number of pitch classes found in the scale B) Pitch (i) Sparse scale: <4 pitch classes (ii) Moderately dense scale: 4–5 pitch classes 8) TONALITY* (iii) Dense scale: >5 pitch classes Organization of discrete pitches around one or (n/a) A-/hetero-/poly-tonal: See (8) more tonic notes N.B. See Box 1 for the definition of a “pitch (a) Indeterminate a-tonal: No discrete pitches class.” (e.g., exclamations, heightened speech) Comments: The more common term “pitch (b) Discrete a-tonal: Discrete pitches, but no tonic class” is used to refer to pitches that share the same (c) Hetero-tonal: Tonic modulates/shifts between note-name regardless of octave, rather than phrases Kolinski’s (1961) term “tint” or the alternative term (d) Poly-tonal: Multiple, simultaneous tonics in “scale degree.” different vocal parts (e) Iso-tonal: Single tonic throughout 11) HEMITONICITY* N.B. See Box 1 for the definition of “tonic.” Percentage of melodic intervals that are Songs not classified as (e) (“iso-tonal”) must be semitones (50–150 cent intervals) coded (n/a) for characters (9–10). (i) Anhemitonic: <5% Comment: Kolinski’s (1961) scale- (ii) Moderately hemitonic: 5–20% classification scheme did not recognize the fact that (iii) Highly hemitonic: >20% some songs have no tonic or have multiple tonics. Comments: Scales are commonly described as Therefore, we have added this character to permit being hemitonic (containing semitones) or classification of these additional types of songs. anhemitonic (not containing semitones). However, this dichotomy fails to recognize the importance of 9) MODE* different gradations in the frequency with which Presence of pitch classes at a minor 3rd (250– semitones are used. 350 cents) or major 3rd (350–450 cents) above the tonic 12) MELODIC INTERVAL SIZE (Cantometrics (a) A-modal: No 3rd present Line 21) (b) Hetero-modal: Both major and minor 3rd Maximum pitch distance between successive appear but in separate phrases notes within any vocal part (c) Poly-modal: Both major and minor 3rd appear (i) Small intervals: <350 cents (i.e., minor 3rd or in the same phrase less; formerly divided into (d) Minor iso-modal: Minor 3rd only “monotone,” “narrow,” and “diatonic” intervals) (e) Major iso-modal: Major 3rd only (ii) Medium intervals: 350–750 cents (i.e., major (n/a) A-/hetero-/poly-tonal: See (8) 3rd – perfect 5th; formerly N.B. See Box 1 for the definition of a “pitch divided into “wide” and “very wide” intervals) class.” (iii) Large intervals*: >750 cents (i.e., minor 6th or Comments: The concept of mode is complex, greater) and the distinction between major and minor 3rd is only one of many possible angles from which to 11

CantoCore: A New Cross-Cultural Song Classification Scheme N.B. Intervals between the final note of a number of contours will end up being classified as phrase and the first note of the next phrase are not “undulating,” reducing the overall informativeness coded. of the character. Comments: Vague definitions from Comments: Although most phrases in both Cantometrics combining interval frequency and humans and birds tend to descend in their final half size, such as “intervals of a half step or less are (Huron 2006; Tierney, Russo and Patel 2011), three prominent (though not necessarily dominant),” were additional character-states were needed to allow for redefined solely in terms of maximum size. A new horizontal, ascending, or U-shaped contours that character-state was created to recognize the were not classifiable by Cantometrics. Cantometrics importance of much larger intervals, such as the Line 19 (“Position of the final tone”) was removed octave. because it was redundant with this character.

13) MELODIC RANGE (Cantometrics Line 20) C) Syllable Maximum pitch distance between the highest and lowest notes within any vocal part 15) MELISMA (Cantometrics Line 29) (i) Small range: <750 cents (i.e., perfect 5th or less) Maximum number of consecutive notes without (ii) Medium range: 750–1250 cents (i.e., perfect 5th articulating a new syllable – octave) (i) Syllabic: 1–2 notes (iii) Large range: >1250 cents (i.e., more than an (ii) Mildly melismatic: 3–5 notes octave) (iii) Strongly melismatic: >5 notes Comments: This character is essentially Comment: While Cantometrics defined unchanged from the 1976 version of Cantometrics. melisma in terms of the frequency of melisma, the current character is defined in terms of maximum 14) MELODIC CONTOUR (Cantometrics Line length to be more consistent with other quantitative 15) characters. Shape resulting from all changes in interval direction within a vocal part 16) VOCABLES (Cantometrics Line 10) (a) Horizontal*: No ascending or descending The percentage of syllables containing only intervals vowels and/or semi-vowels (e.g., “y,” “h,” “w”) (b) Ascending*: Ascending intervals only (i) Few vocables: <20% (formerly “little or no (c) Descending: Descending intervals only repetition”) (formerly divided into “descending” and (ii) Some vocables: 20-50% (formerly divided into “terraced” contours) “some repetition” and “half repetition”) (d) U-shaped*: First descending, then ascending (iii) Many vocables: >50% (formerly divided into intervals “quite repetitious” and “extreme repetition”) (e) Arched: First ascending, then descending Comment: Vocables (non-lexical “nonsense” intervals syllables) are an important feature of many musics (f) Undulating: Multiple changes of interval cross-culturally but are difficult to define and code direction for someone who does not speak the language N.B. Each phrase should be treated as having its (Maranda 1970). This led Lomax to change the own contour, except when there are clear “hyper- emphasis from vocables to textual repetition, but phrase” contours that connect multiple phrases. this change of emphasis becomes confounded with Cases where multiple contours appear in different phrase repetition (21). The current character instead phrases and/or different vocal parts should be multi- uses words containing only vowels and/or semi- coded. Some discretion must be used in deciding vowels as a proxy for vocables. If the coder what constitutes a change of interval direction. In understands the language, they may exclude lexical general, temporary interval changes that do not semi-vowel/vowel combinations (e.g., “yo-yo” in greatly affect the dominant melodic contour should English) and/or include vocables that include be ignored (e.g., changes of interval direction that consonants (e.g., Celtic mouth music). last only one or two notes). Otherwise, a large 12 Analytical Approaches To World Music 2.1 (2012) 87-137 II) “TEXTURE” (between-part) because they are commonly used to categorize texture as a whole, despite ambiguities about 17) NUMBER OF VOCAL PARTS distinguishing between rhythmic texture, harmonic (Cantometrics Line 4) texture, and relative motion. Maximum number of simultaneous vocal parts This character concerns the rhythmic (i) One-part: 1 (formerly divided into “solo” and relationship between the notes of multiple parts, “unison”) regardless of what meter those parts are in. Thus, (ii) Two-part*: 2 “iso-metric” songs (see 1) with rhythmically (iii) Many-part*: >2 independent parts can be “poly-rhythmic” despite N.B. See Box 1 for the definition of a “vocal not being “poly-metric.” The two types of poly- part.” Songs classified as (a) (“one-part”), including meter that were originally also classified in this solo, unison, and doubling at the octave (i.e., character have been moved to (1). “magadizing”), must be coded (n/a) for characters (18–20). This does not include “multisonance” 19) HARMONIC TEXTURE* (multiple pitch classes realized simultaneously Minimum harmonic interval (octave- [Kolinski 1978]), whether intentional or not. equalized—see N.B. below) between “Many-part” songs may require multi-coding for simultaneous vocal parts that is sustained for at characters (18–20), as may “two-part” songs that least 1 second transition between different texture types. (i) Rough (“dissonant”): 50–249 cents (includes Comments: This character no longer 951–1150 cents) (e.g., 2nds/7ths) distinguishes between the number of voices singing (ii) Smooth (“consonant”): 250–600 cents (includes each part or their rhythmic relationship, which are 600–950 cents) (e.g., 3rds/6ths) now coded in (24) and (18), respectively. Further (n/a) One-part (includes 0–49 and 1150–1200 distinctions between the numbers of parts (e.g., 3- cents), or poly-/hetero-rhythmic: See Box 1 and part, 4-part, 5-part, etc.) were avoided because it (17/18) proved difficult to reliably code beyond three parts. N.B. Harmonic intervals should be calculated after correcting for absolute differences in pitches 18) RHYTHMIC TEXTURE (Cantometrics Line by transposing them to the octave that minimizes 12) the harmonic interval. For example, the top note in a Temporal asynchrony in the relative onsets of harmonic interval of 1000 cents (minor 7th) can be different vocal parts (in seconds) transposed down one octave to create a harmonic (a) Hetero-rhythmic (heterophonic): 0.1–1s interval of 200 cents (major 2nd). Therefore, the (formerly “rhythmic heterophony”) largest possible harmonic interval is 600 cents (a (b) Poly-rhythmic (polyphonic): >1s (formerly tritone) before the octave-equalized interval size divided into “accompanying rhythm” and begins to decrease again. “rhythmic counterpoint”) Comments: While most quantitative characters (c) Iso-rhythmic (homophonic): <0.1s (formerly are defined in terms of maximum values, this “rhythmic unison”) character is defined in terms of minimum values (n/a) One-part (monophonic): See (17) because most songs with rough intervals also N.B. Unison songs where multiple singers contain smooth intervals, but not the reverse. To generally sing the same pitches with less than 0.1s prevent confusion, we avoid the related terms offset are classified as “one-part,” not “iso- “dissonant” and “consonant” as well as common rhythmic” (see Box 1 and 17). Songs with different distinctions between “consonant,” “perfect” and rhythmic textures between different vocal parts or “tritone” intervals found in Western music theory, different phrases should be multi-coded. Songs not as it is not yet established to what degree these classified as “iso-rhythmic” must be coded “n/a” for categories are cross-culturally or experimentally character (19). valid. Plomp and Levelt (1965) developed an Comments: The corresponding terms “poly- experimentally-based explanation for sensory /hetero-/homo-/mono-phonic” have been included consonance based on the acoustic critical 13

CantoCore: A New Cross-Cultural Song Classification Scheme bandwidth, but there also alternative usage-oriented listed in an order where the number of phrases definitions. Although calculating critical bandwidth decreases rather than increases, just as they were in is impractical to do by ear, it basically corresponds Cantometrics. The original Cantometric character to our “rough/smooth” division as well as to the contained 13 different character-states, each with a traditional division in Western music theory specific combination of features (e.g., “complex between “consonant” 3rds/6ths and “dissonant” strophe with little/no variation,” “simple litany with 2nds/7ths. It should be noted that the critical high variation,” etc.). Busby (2006) reorganized bandwidth is more complex and nuanced, varying this character into three new characters— “phrase throughout the audible range and giving different repetition,” “complexity,” and “amount of sensory consonance values for 2nds vs. 7ths. variation”—but we retained only the “phrase repetition” character, as we found the other two 20) RELATIVE MOTION (Cantometrics Line characters too difficult to reliably define and code. 22) “Canonic/round form” and other overlapping Relationship of the melodic contours (see 13) relationships between parts are now coded in (26). of two simultaneous parts (a) Hetero-contour (drone): One part is horizontal, 22) PHRASE LENGTH (Cantometrics Line 17) the other changes direction (formerly “drone Maximum phrase length, in seconds polyphony”) (i) Short phrases: <5 s (formerly divided into “very (b) Poly-contour (independent motion): Both parts short” and “short” phrases) have different, non-horizontal contours (ii) Medium-length phrases: 5–9 s (formerly divided into “harmony” and (iii) Long phrases: >9 s (formerly divided into “counterpoint”) “long” and “very long” phrases) (c) Iso-contour (parallel motion): Both parts have Comments: As stated previously, ambiguities the same contour (formerly divided into about where a phrase ends should be resolved by “isolated chords” and “parallel chords”) relying on breathing points to define phrase (n/a) One-part: See (17) boundaries. N.B. Songs with different types of relative motion between different vocal parts or different 23) PHRASE SYMMETRY (Cantometrics Line phrases should be multi-coded. 18) Comments: Distinctions between different Ratio of the length of the longest phrase in a “poly-contour” and “iso-contour” sub-types song relative to the shortest phrase (including ostinato) were removed due to their (i) Symmetric: <1.5 times the length of the shortest vague definitions. phrase (ii) Mildly asymmetric*: 1.5–2.5 times the length of III) “FORM” (between-phrase) the shortest phrase (iii) Very asymmetric*: >2.5 times the length of the 21) PHRASE REPETITION (Cantometrics Line shortest phrase 16) Comments: The original character did not Maximum number of successive phrases before define “symmetry.” Characters in the original a phrase is repeated character regarding the number of phrases were (i) Non-repetitive: >8 phrases, or no repeat at all removed because they were redundant with phrase (formerly “through-composed”) repetition (21). (ii) Moderately repetitive: 3–8 phrases (formerly “strophe”) 24) SOLO/GROUP ARRANGEMENT (iii) Repetitive: 1–2 phrases (formerly “litany”) (reorganization of Cantometrics Line 1) N.B. See Box 1 for the definition of a “phrase.” Number of singers in each phrase Phrases where everything but the text is repeated (a) Solo: Only solo phrases throughout (formerly are counted as a repeat for this character. divided into “one solo singer” and “one solo Comments: Because of the way phrase singer after another”) repetition is operationalized, the character-states are 14 Analytical Approaches To World Music 2.1 (2012) 87-137 (b) Mixed: Individual phrases contain both group 26) PHRASE OVERLAP (reorganization of and solo sub-sections (formerly “social unison Cantometrics Line 1) with a dominant leader”) Maximum overlap between a “call” phrase and (c) Alternating: Alternation between distinct solo the “response” phrase that alternates with it (as and group phrases (formerly divided into the percentage of time in which the latter “simple alternation: leader-chorus,” phrase overlaps with the former) “overlapping alternation: leader-chorus,” and (i) Non-overlapping: 0% (formerly divided into “overlapping alternation: chorus-leader”) “simple alternation: chorus-chorus,” “simple (d) Group: Only group phrases throughout alternation: leader-chorus,” and “one solo (formerly divided into “social unison with the singer after another”) group dominant,” “discoordinated,” “simple (ii) Mildly overlapping: 1–25% (formerly divided alternation: chorus-chorus,” “overlapping into “overlapping alternation: chorus-chorus” alternation: chorus-chorus,” and “interlock”) and “overlapping alternation: chorus-leader”) Comments: Cantometrics Line 1 originally (iii) Highly overlapping: >25% (formerly classified contained 13 character-states that represented as “interlock” and/or “canonic or round form” various complex combinations of multiple in Line 16) characters. Characters involving solo/group (n/a) A-responsorial: See (25) arrangement, responsorial arrangement, and phrase Comments: See comments in (24). overlap have been moved to characters (24), (25), and (26), respectively, to isolate the common Sample classification features in these underlying characters (following Busby 2006). To aid in understanding the practicalities involved in applying these idealized definitions to 25) RESPONSORIAL ARRANGEMENT real songs, a sample transcription of the Shona song (reorganization of Cantometrics Line 1) “Pi mcinanga” (track 13 from Lomax’s [1976] Alternation of phrases between different vocal Cantometrics Consensus Tape) is provided along parts with a table showing how it would be classified (a) A-responsorial: No alternation between parts (Figure 4). CantoCore classifications for all 30 (formerly divided into “one solo singer,” songs on the Cantometrics Consensus Tape are “social unison with the group dominant,” listed in Appendix A. “discoordinated,” and “social unison with a dominant leader”) RELIABILITY (b) Hetero-responsorial*: Irregular alternation To compare the inter-rater reliability of each between parts system, E.M. used both Cantometrics and (c) Iso-responsorial: Consistent alternation CantoCore, to classify the 30 songs from the between parts (formerly divided into “simple Cantometrics Consensus Tape (Figure 5) by ear alternation: chorus-chorus,” “overlapping after being trained in both systems with the aid of alternation: chorus-chorus,” “simple the Cantometrics Training Tapes (Lomax 1976), but alternation: leader-chorus,” “overlapping before being informed of our hypotheses about alternation: chorus-leader,” “one solo singer reliability. We then compared her Cantometric after another,” and “interlock”) codings with those of the creators of Cantometrics N.B. Songs classified as (a) (“a-responsorial”) (Lomax 1976, 168–70) and her CantoCore codings must be coded (n/a) for character (26). with those of one of its creators (P.E.S.; his codings Comments: See comments in (24). are shown in Appendix A). We calculated the agreement on each character separately (see Appendix B tables B1–3 for detailed results), and then averaged across all characters to compare the

15

CantoCore: A New Cross-Cultural Song Classification Scheme

Character Quantitative value Classification 1) Meter n/a d) Iso-metric 2) No. of beats n/a b) Triple 3) Beat sub-division n/a c) Iso-divisive 4) No. of sub-beats n/a a) Simple 5) Syncopation 2% (1/61 notes [2nd note of bar 9]) i) Un-syncopated

6) Motivic redundancy 66% (40/61 notes derived from e r r e) iii) Highly motivic

7) Durational variability 3 unique duration values (r, e , and e. ) ii) Moderate durational variability 8) Tonality n/a e) Iso-tonal 9) Mode n/a d) Minor iso-modal 10) Number of pitch classes 6 pitch classes (A,B,C,D,E ,G) iii) Dense scale 11) Hemitonicity 7% (4/60 intervals are the semitone ii) Moderately hemitonic between C and B) 12) Melodic interval size 800 cents max [C-E in bars 9–10] iii) Large intervals 13) Melodic range 1200 cents [E-E] ii) Medium range 14) Melodic contour n/a cef) Descending [phrases 2, 4 & 5]; arched [phrases 6 & 8]; undulating [phrases 1,3 & 7] 15) Melisma 1 note max i) Syllabic 16) Vocables 13% [8/61 syllables] i) Few vocables 17) No. of vocal parts 1 i) One-part 18) Rhythmic texture n/a n/a) One-part 19) Harmonic texture n/a n/a) One-part 20) Relative motion n/a n/a) One-part 21) Phrase repetition Max. of 3 new phrases (phrases 5–7) are ii) Moderately repetitive introduced before an earlier phrase (phrase 6) is repeated 22) Phrase length 2 seconds max i) Short phrases 23) Phrase symmetry 1 (1:1 ratio of longest:shortest phrase) i) Symmetric 24) Solo/group arrangement n/a c) Alternating 25) Responsorial arrangement n/a c) Iso-responsorial 26) Phrase overlap 0% i) Non-overlapping Figure 4. Transcription of the Shona song “Pi mcinanga” (track 13 from the Cantometrics Consensus Tape [Lomax 1976]) and its codings on the 26 CantoCore characters. Phrases (all are two measures long) are shown using phrase marks. Syllables containing only vowels and/or semi-vowels (used as a proxy for vocables) are underlined. The actual pitches are two semitones lower than those shown in the transcription. For quantitative characters, both raw quantitative values and categorical classifications are shown. An mp3 file is available at http://greenstone.ilam.ru.ac.za/collect/ilam/index/assoc/D11849.dir/TR174-09.mp3 mean agreement between the two classification agreeing that a song is sung solo inflates the true systems We calculated inter-rater reliability for each agreement because a “solo” character-state is coded individual character in two ways. First, we used the redundantly in six different Cantometric characters simplest measure, that of percent agreement. related to vocal texture and vocal blend. Therefore, However, this statistic does not account for the we also calculated reliability a second way, this effects of chance agreement, partial agreement, and time correcting for these problems using the kappa- character redundancy. For example, some amount statistic (κ), after removing all redundant codings of agreement would be expected by chance even if (i.e., all “n/a” codings in CantoCore and character- the coders were coding at random, but some types state “1” [“absence”] for Cantometrics lines 2, 4–9, of disagreement (e.g., “good blend” vs. “maximal 12–14, 22, and 27). We used “weighted κ” (squared blend”) are less severe than others (e.g., “maximal weighting) (Cohen 1968) for quantitative characters blend” vs. “no blend (solo)”). Furthermore, simply and “unweighted κ” (Cohen 1960) for qualitative 16 Analytical Approaches To World Music 2.1 (2012) 87-137

Figure 5. Approximate geographic locations for the 30 songs from the Cantometrics Consensus Tape (Lomax 1976) used to test the reliability of Cantometrics and CantoCore. The map was generated using the World Atlas of Language Structures Online (http://wals.info). characters. For both percent agreement and κ, we partial agreement (including both the observed used only the single coding indicated as most partial agreement and the joint marginal probability prominent in cases of multi-coding. of this agreement), we arrive at a final weighted κ For example, for Cantometrics Line 5 (tonal value of 0.49. blend of the vocal part), the coders agreed on 21 out The results for κ are shown in Figure 6. As of the 30 codings, giving a percent agreement value predicted, CantoCore appeared to be more reliable of 70%. However, 10 of these cases simply than Cantometrics. The mean percent agreement represented a repetition of an agreement that there was 62% for CantoCore and 45% for Cantometrics. was only one vocal part, a character-state that had The results using the κ statistic were highly already been coded in Line 4. When we limit this significant statistically (p=0.0001), with the mean κ character’s analysis to only the 18 characters where value of CantoCore (0.47) being approximately both coders had already agreed in Line 4 that there 80% higher than that of Cantometrics (0.26). were in fact multiple voices to base an estimation of According to Landis and Koch’s (1977) criteria for tonal blend upon, the percent agreement value interpreting κ, this translates to “moderate” would be 61%. However, this does not account for reliability for CantoCore and “fair” reliability for the amount of chance agreement we would expect Cantometrics, on a scale of “poor” (<0), “slight” (0– given each coder’s baseline propensity for choosing 0.2), “fair” (0.21–0.4), “moderate” (0.41–0.6), each character-state (the “joint marginal “substantial” (0.61–0.8), and “almost perfect” probability”), which in this case is 31%. Cohen’s (0.81–1). Both systems were significantly more Kappa effectively calculates the proportion of reliable than chance (p<1x10-11), countering claims agreement after subtracting out this chance that Cantometrics is unreliable (Downey 1970; agreement as follows: κ = (0.61–0.31) / (1–0.31) = Maranda 1970; Nettl 1970). 0.43. However, this still does not account for the There is some debate about how to interpret degree of partial agreement in cases where the kappa-statistics, as Landis & Koch’s criteria, coders did not exactly agree. For example, “4) good although useful, are self-admittedly arbitrary. Some blend” is three times closer to “5) maximal blend” authors have proposed further additions to κ, such than is “2) no blend [solo].” With squared as using the maximum attainable κ given the weighting, disagreement between 5 and 2 is coders’ pre-existing marginal probabilities or using weighted as 32, i.e., nine times as severe as a minimum acceptable threshold value for κ (e.g., disagreement between 4 and 5. Once we also 0.4 for clinical uses) rather than zero (Sim & Wright incorporate information about degree of weighted 2005). However, it should be noted that Cohen 17

CantoCore: A New Cross-Cultural Song Classification Scheme All of the reliability values for both CantoCore and Cantometrics are substantially lower than the ones given by Lomax (1976, 270) and by Lomax, Halifax and Markel (1968). However, it is difficult to compare these datasets with our results, as they used different statistics, did not present complete data or methods, and did not use a consistent song sample. At the same time, our own data should be treated as provisional, as logistical constraints limited us to collecting reliability data from only a single coder. Victor Grauer (personal communication) has pointed out that our results may be a stronger reflection on our coder and/or our training procedure than on the classification schemes themselves. We accept this possibility but maintain that we have tried our best not to bias the test in favor of CantoCore. We therefore predict that Figure 6. Mean reliability for all 37 classification characters the relative reliability values of the two schemes in Cantometrics and all 26 characters in CantoCore. Error bars will probably remain similar even if the absolute represent the standard error of the mean. CantoCore is reliability values for both schemes is higher or significantly more reliable than Cantometrics (p=0.0001). lower overall for different coders. Of course, as with all science, our claims should be tested by (1960) originally advised against giving much independent researchers with larger samples to see weight to maximum attainable κ, as “disagreement whether they are replicable and whether they can which is forced by marginal disagreement has the generalize to other situations and other cultures. same negative consequences as that not so forced— One possibility to improve reliability for both in short, it is disagreement,” and that a minimum systems in the future would be to have multiple acceptable threshold value of 0.4 is equally as independent coders classify songs and construct a arbitrary as Landis & Koch’s criteria. consensus coding based on their combined Contrary to our predictions, there was no agreement. significant difference in reliability between the structural and performance characters of Cantometrics (structure: mean κ = 0.30, APPLICATIONS performance: mean κ = 0.29; p=0.81) (Tables B1– Classification is a method of examining patterns 2). Therefore, it may still be useful to supplement of similarity and difference. It is a means, not an CantoCore’s structural characters with the end. Thus, the true test of CantoCore will be performance characters from Cantometrics. whether it, like Cantometrics, can be used as a tool Cantometrics’ instrumentation characters (Table to explore relationships between songs, and between B3), however, were almost three times less reliable music and culture. Comparing the relative than its structural and performance characters similarities and differences across all CantoCore (instrumentation: mean κ = 0.11). This is consistent classifications can allow us to quantify different with our prediction that songs are more amenable degrees and types of musical similarity. This can be than instrumental music to reliable cross-cultural used to create global musical taxonomies, in the classification. We could not reject the null same way that Cantometrics permitted Lomax hypothesis that agreement on instrumental (1968) to propose 10 canonical singing styles characters was simply due to chance at the standard throughout world cultures. significance threshold of p=0.05, although the The growth of the digital humanities has seen obtained value of p=0.07 is so close to this the birth of a new field of Music Information threshold that it may well have been significant Retrieval (MIR) eager to take up the challenge of given a larger sample size. classifying music. While MIR has made great 18 Analytical Approaches To World Music 2.1 (2012) 87-137 strides in adapting computational models to The time has come to return to Adler’s (1885) Western music, the lack of a theoretical framework original vision of a musicology that sees for cross-cultural musical classification still classification, comparison, and ethnography as hampers the development of “computational equal partners in the quest to understand the world ethnomusicology” (Tzanetakis et al. 2007). Both of music. This will require using all the tools that CantoCore and Cantometrics provide such a are available, musical and non-musical, humanistic framework, which computational and scientific, qualitative and quantitative, ethnomusicologists can build on to design theoretical and empirical. It will also require automated algorithms to allow for faster and more collaborative approaches that integrate work objective classification and acoustic feature- ranging from “thick description” (Geertz 1973) of extraction. Our classification of the 30 songs from individual songs or societies to “mass comparison” the Cantometrics Consensus Tape can act as a (Greenberg 1957) of worldwide patterns of “ground-truth” dataset for such attempts or for diversity. Anthropologists have historically been atheoretical classification approaches using split between those in the humanities who statistically based machine-learning algorithms. emphasize the former and those in the sciences who Furthermore, our improved statistical techniques for emphasize the latter, but there has recently been a examining song similarity (Rzeszutek, Savage and movement towards integrating both approaches Brown 2012) provide new methods that (Kuper and Marks 2011; Smith, Gurven, and computational ethnomusicologists can use to Mulder 2011; Nekaris, Nijman, and Godfrey 2011). analyze similarity not only between individual CantoCore provides a reliable method to assist in songs but also between diverse repertoires of this multidisciplinary goal. heterogeneous songs. Classification is also a tool that can be used to ACKNOWLEDGMENTS provide insight into musical evolution and human We are grateful to Geoffrey Clarfield, Tom history. While much of the study of musical Currie, Victor Grauer, Joseph Jordania, David evolution has focused on music’s role in biological Locke, Michael Tenzer, and Anna Lomax Wood for evolution (Spencer 1857; Darwin 1871; Pinker critical reading of a previous version of the 1997; Wallin, Merker, and Brown 2000; Cross manuscript. We thank Sawa Matsueda Savage for 2001), little attention has been given to the cultural the art work in Figure 1. This work was supported evolution of music itself, including the forces of by the Social Sciences and Humanities Research musical change and stasis both geographically and Council (SSRHC) of Canada to S.B. and by an historically. When attempts have been made in this Amherst College Roland Wood Fellowship and a direction (Lomax 1968; Lomax and Berkowitz Japanese Ministry of Education, Science, Sports and 1972; Grauer 2006; Jan 2007), critics have rightly Technology Scholarship to P.E.S. pointed out difficulties in distinguishing “deep”

(phylogenetic) evolutionary relationships from REFERENCES “surface” (phenetic) acoustic similarities (Blacking 1977; Stock 2006). However, similar issues also Adler, Guido. 1885. “The Scope, Method and Aim confront the study of biological evolution (Hennig of Musicology [in German]. ”Vierteljahrsschrift 1965; Sneath and Sokal 1973; Doolittle 1999) and fur Musikwissenschaft 1: 5–20. cultural evolution (Mace and Holden 2005; Currie, Agawu, Kofi. 2010. “Against Ethnotheory.” Paper Greenhill, and Mace 2010). Importantly, presented at the First International Conference classification tools like Cantometrics and on Analytic Approaches to World Music, CantoCore provide a typological view of music— Amherst, Massachusetts, February 19–21. breaking music down into the principal characters Anku, Willie. 2000. “Circles and Time: A Theory of that make up these schemes—and this may be Structural Organization of Rhythm in African useful in understanding the evolution of individual Music.” Music Theory Online 6.1. ISBN:1067- musical characters as well as elucidating musical 3040. universals (Brown and Jordania 2011). 19

CantoCore: A New Cross-Cultural Song Classification Scheme Atkinson, Quentin D. 2011. “Phonemic Diversity of Favoured Races in the Struggle for Life. Supports a Serial Founder Effect Model of London: John Murray. Language Expansion from Africa.” Science 332: ______. 1871. The Descent of Man, and Selection in 346–349. Relation to Sex. London: John Murray. Bertin-Mahieux, Thierry, Dan P.W. Ellis, Brian Doolittle, W. Ford. 1999. “Phylogenetic Whitman, and Paul Lamere. 2011. “The Million Classification and the Universal Tree.” Science Song Dataset.” Proceedings of the 12th 284.5423: 2124–2128. International Society for Music Information Dowling, W. Jay, and Dane L. Harwood. 1986. Retrieval Conference (ISMIR 2011): 591–596. Music Cognition. Orlando: Academic Press. Blacking, John. 1973. How Musical is Man? Downey, James C. 1970. “Review of A. Lomax, Seattle: University of Washington Press. Folk Song Style and Culture.” Ethnomusicology ______. 1977. “Some Problems of Theory and 14.1: 63–67. Method in the Study of Musical Change.” Driver, Harold E. 1970. “Review of A. Lomax, Folk Yearbook of the International Folk Music Song Style and Culture.” Ethnomusicology 14.1: Council 9: 1–26. 57–62. Brown, Steven. 2000. “The ‘Musilanguage’ Model Dunn, Michael, Simon J. Greenhill, Stephen C. of Musical Evolution.” In The Origins of Music, Levinson, and Russell D. Gray. 2011. “Evolved ed. Nils L. Wallin, Bjorn Merker, and Steven Structure of Language Shows Lineage-Specific Brown, 271–300. Cambridge, MA: MIT Press. Trends in Word-Order Universals.” Nature 473: Brown, Steven, and Joseph Jordania. 2011. 79–82. “Universals in the World’s Musics.” Psychology Ellis, Alexander J. 1885. “On the Musical Scales of of Music. doi:10.1177/0305735611425896. Various Nations.” Journal of the Society of Arts Busby, George. 2006. “Finding the Blues: An 33.1: 485–527. Investigation into the Origins and Evolution of Erickson, Edwin E. 1976. “Tradition and Evolution African-American Music.” M.Sc. thesis: in Song Style: A Reanalysis of Cantometric University of London. Data.” Cross-Cultural Research 11.4: 277–308. Cohen, Jacob. 1960. “A Coefficient of Agreement Feld, Steven, and Aaron A. Fox. 1994. “Music and for Nominal Scales.” Educational and Language.” Annual Review of Anthropology 23: Psychological Measurement 20.1: 37–46. 25–53. ______. 1968. “Weighted Kappa: Nominal Scale Geertz, Clifford. 1973. The Interpretation of Agreement with Provision for Scaled Cultures. New York: Basic Books. Disagreement or Partial Credit.” Psychological Grauer, Victor. 2005. “‘Cantometrics—Song and Bulletin 70.4: 213–220. Social Culture’: A Response.” Musical Cross, Ian. 2001. “Music, Cognition, Culture, and Traditions 159. http://www.mustrad.org.uk/ Evolution.” Annals of the New York Academy articles/cantome2.htm. of Sciences 930: 28–42. Grauer, Victor. 2006. “Echoes of our Forgotten Currie, Thomas E., Simon J. Greenhill, and Ruth Ancestors.” The World of Music 48.2: 5–59. Mace. 2010. “Is Horizontal Transmission Greenberg, Joseph H. 1957. Essays in Linguistics. Really a Problem for Phylogenetic Comparative Chicago: University of Chicago Press. Methods? A Simulation Study Using Harris, Marvin. 1976. “History and Significance of Continuous Cultural Traits.” Philosophical the Emic/Etic Distinction.” Annual Review of Transactions of the Royal Society B: Biological Anthropology 5.1: 329–350. Sciences 365.1559: 3903–3912. Haspelmath, Martin, Matthew S. Dryer, David Gil, Currie, Thomas E., and Ruth Mace. 2009. “Political and Bernard Comrie, eds. 2005. The World Complexity Predicts the Spread of Atlas of Language Structures. Oxford: Oxford Ethnolinguistic Groups.” Proceedings of the University Press. National Academy of Sciences of the United Hennig, Willi. 1965. “Phylogenetic Systematics.” States of America 106.18: 7339–7344. Annual Review of Entomology 10.1: 97–116. Darwin, Charles. 1859. On the Origin of Species by Means of Natural Selection, or the Preservation 20 Analytical Approaches To World Music 2.1 (2012) 87-137 Henry, Edward O. 1976. “The Variety of Music in a Lomax, Alan, ed. 1968. Folk Song Style and North Indian Village: Reassessing Culture. Washington, DC: American Cantometrics.” Ethnomusicology 20.1: 49–66. Association for the Advancement of Science. Hood, Mantle. 1971. The Ethnomusicologist. New Lomax, Alan, and Norman Berkowitz. 1972. “The York: McGraw-Hill. Evolutionary Taxonomy of Culture.” Science Hornbostel, Erich M. von, and Curt Sachs. 1914. 177.4045: 228–239. “Classification of Musical Instruments [in Lomax, Alan, and Victor Grauer. 1968. “The German].” Zeitschrift fur Ethnologie 14: 3–29. Cantometric Coding Book.” In Folk Song Style Huron, David. 2006. Sweet Anticipation: Music and and Culture, ed. Alan Lomax, 34–74. the Psychology of Expectation. Cambridge: MIT Washington, DC: American Association for the Press. Advancement of Science. Jan, Steven. 2007. The Memetics of Music: A Neo- Lomax, Alan, Joan Halifax, and Norman N. Markel. Darwinian View of Musical Structure and 1968. “Consensus on Cantometric Parameters.” Culture. Hants: Ashgate. In Folk Song Style and Culture, ed. Alan Jones, William. 1807. The Works of Sir William Lomax, 111–116. Washington, DC: American Jones. 1st ed. London: John Hatchard. Association for the Advancement of Science. Kolinski, Mieczyslaw. 1961. “Classification of London, Justin. 1995. “Some Examples of Complex Tonal Structures.” Studies in Ethnomusicology Meters and their Implications for Models of 1: 38–76. Metric Perception.” Music Perception 13.1: 59– ______. 1962. “Consonance and Dissonance.” 77. Ethnomusicology 6.2: 66–74. ———. 2004. Hearing in Time: Psychological ______. 1973. “A Cross-Cultural Approach to Aspects of Musical Meter. Oxford: Oxford Metro-Rhythmic Patterns.” Ethnomusicology University Press. 17.3: 494–506. Mace, Ruth, and Clare J. Holden. 2005. “A ______. 1978. “The Structure of Music: Phylogenetic Approach to Cultural Evolution.” Diversification Versus Constraint.” Trends in Ecology and Evolution 20.3: 116–21. Ethnomusicology 22.2: 229–244. Maranda, Elli Kongas. 1970. “Deep Significance Krumhansl, Carol. 1990. Cognitive Foundations of and Surface Significance: Is Cantometrics Musical Pitch. New York: Oxford University Possible?” Semiotica 2.2: 173–184. Press. McLeod, Norma. 1974. “Ethnomusicological Kuper, Adam, and Jonathan Marks. 2011. Research and Anthropology.” Annual Review of “Anthropologists Unite!” Nature 470.7333: Anthropology 3.1: 99–115. 166–168. Merriam, Alan P. 1982. “On Objections to Landis, J. Richard, and Gary G. Koch. 1977. “The Comparison in Ethnomusicology.” In Cross- Measurement of Observer Agreement for cultural Perspectives on Music, ed. Robert Categorical Data.” Biometrics 33.1: 159–174. Falck and Timothy Rice, 175–189. Toronto: Lerdahl, Fred, and Ray Jackendoff. 1983. A University of Toronto Press. Generative Theory of Tonal Music. Cambridge, Naroll, Raoul. 1969. “Singing and Dancing: A MA: MIT Press. Cross-Cultural Survey.” Science 166.3903: 366– Leroi, Armand M., and Jonathan Swire. 2006. “The 367. Recovery of the Past.” The World of Music 48.3: Nekaris, K. Anne-Isola, Vincent Nijman, and Laurie 43–54. R. Godfrey. 2011. “Anthropology: Follow Field Lewis, M. Paul, ed. 2009. Ethnologue: Languages Primatologists.” Nature 471: 448. of the World. 16th ed. Dallas: SIL International. Nettl, Bruno. 1970. “Review of A. Lomax, Folk Lomax, Alan. 1959. “Folk Song Style.” American Song Style and Culture.” American Anthropologist 61.6: 927–954. Anthropologist 72.2: 438–441. ______. 1976. Cantometrics: An Approach to the ______. 2005. The Study of Ethnomusicology: Anthropology of Music. Berkeley: University Thirty-One Issues and Concepts. 2nd ed. of California Extension Media Center. Champaign: University of Illinois Press. 21

CantoCore: A New Cross-Cultural Song Classification Scheme Patel, Aniruddh D. 2008. Music, Language and the Interpretation, and Sample Size Requirements.” Brain. Oxford: Oxford University Press. Physical Therapy 85.3: 257–268. Patel, Aniruddh D., and Joseph R. Daniele. 2003. Smith, Eric Alden, Michael Gurven, and Monique “An Empirical Comparison of Rhythm in Borgerhoff Mulder. 2011. “Anthropology: It Language and Music.” Cognition 87.1: 35–45. Can Be Interdisciplinary.” Nature 471: 448. Perlman, Marc, and Carol L. Krumhansl. 1996. “An Sneath, Peter H.A., and Robert R. Sokal. 1973. Experimental Study of Internal Interval Numerical Taxonomy: The Principles and Standards in Javanese and Western Musicians.” Practice of Numerical Classification. San Music Perception 14.2: 95–116. Francisco: W.H. Freeman and Company. Pfordresher, Peter Q., Steven Brown, Kimberly M. Spencer, Herbert. 1857. “The Origin and Function Meier, Michel Belyk, and Mario Liotti. 2010. of Music.” Fraser’s Magazine 56: 396–408. “Imprecise Singing is Widespread.” The Journal Stock, Jonathan P. J. 2006. “Clues from our Present of the Acoustical Society of America 128.4: Peers? A Response to Victor Grauer.” The 2182–2190. World of Music 48.2: 73–91. Pinker, Steven. 1997. How the Mind Works. New Stone, Ruth, Bruno Nettl, James Porter, and Tim York: Norton. Rice, eds. 1998. The Garland Encyclopedia of Plomp, R., and W.J.M. Levelt. 1965. “Tonal World Music (10 vols., 9 CDs). New York: Consonance and Critical Bandwidth.” The Routledge. Journal of the Acoustical Society of America 38: Tenzer, Michael, ed. 2006. Analytical Studies in 548–560. World Music. New York: Oxford University Powers, Harold S., Frans Wiering, James Porter, Press. James Cowdery, Richard Widdess, Ruth Davis, Tierney, Adam T., Frank A. Russo, and Aniruddh Marc Perlman, Stephen Jones, and Allan Marett. D. Patel. 2011. “The Motor Origins of Human 2012. “Mode.” Grove Music Online. Oxford and Avian Song Structure.” Proceedings of the Music Online. National Academy of Sciences 108.37: 15510– http://www.oxfordmusiconline.com/subscriber/ 15515. article/grove/music/43718#S43718 Toiviainen, Petri, and Tuomas Eerola. 2001. “A (accessed March 1, 2012). Method for Comparative Analysis of Folk Rzeszutek, Tom, Patrick E. Savage, and Steven Music Based on Musical Feature Extraction and Brown. 2012. “The Structure of Cross-Cultural Neural Networks.” In Proceedings of the VII Musical Diversity.” Proceedings of the Royal International Symposium of Systematic and Society B: Biological Sciences 279: 1606-1612. Comparative Musicology and the III Sachs, Curt. 1943. The Rise of Music in the Ancient International Conference on Cognitive World: East and West. New York: Norton. Musicology, ed. H. Lappalainen, 41–45. Schaffrath, Helmut. 1995. The Essen Folksong Jyväskylä: University of Jyväskylä. Database in Kern Format, ed. D. Huron. Menlo Tzanetakis, George, Ajay Kapur, W. Andrew Park, CA: Center for Computer Assisted Schloss, and Matthew Wright. 2007. Research in the Humanities. “Computational Ethnomusicology.” Journal of Schenker, Heinrich. 1979. Free Composition (Der Interdisciplinary Music Studies 1.2: 1–24. freie Satz): Volume III of New Musical Theories Wallin, Nils L., Bjorn Merker, and Steven Brown, and Fantasies. New York: Longman. eds. 2000. The Origins of Music. Cambridge, Sim, Julius, and Chris C. Wright. 2005. “The Kappa MA: MIT Press. Statistic in Reliability Studies: Use,

22 Analytical Approaches To World Music 2.1 (2012) 87-137 APPENDICES

Appendix A: Sample CantoCore codings Table A1. CantoCore codings of all 30 songs from the Cantometrics Consensus Tape, as done by P.E.S. The 30 songs (see Lomax 1976, 164–171 for details) are listed by row number, and the 26 CantoCore characters are listed by column number. In cases of multi- coding, the most prominent coding is bolded. See text for a detailed description of the characters and character-states. Recordings of these songs are available on Tape VII of Lomax (1976), and are scheduled to be re-released digitally by the Association for Cultural Equity at http://research.culturalequity.org/cls.jsp.

CantoCore character number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 1 ad a c a ii iii iii e ce ii ii ii iii acef i ii iii bc ii abc ii ii iii b b i 2 d a c a i iii ii e e ii i ii iii f i ii i n/a n/a n/a iii ii i a a n/a 3 d a c a ii iii i e e ii ii ii ii ac i iii i n/a n/a n/a ii ii i d a n/a 4 a n/a n/a n/a n/a ii iii e e iii ii ii iii f iii i i n/a n/a n/a i ii i a a n/a 5 d a c a ii iii ii e e ii i iii iii cf i i i n/a n/a n/a ii i i a a n/a 6 d a c a iii iii ii e e ii iii ii iii e iii i i n/a n/a n/a ii ii ii a a n/a 7 a n/a n/a n/a n/a i iii e e ii ii ii i af iii i ii bc i a ii iii iii d a n/a 8 b n/a n/a n/a i ii ii e e ii i iii ii ae iii iii i n/a n/a n/a ii ii ii d a n/a 9 d a c b i iii ii e e iii ii iii iii ef i i i n/a n/a n/a iii ii i a c ii 10 d c b n/a ii iii ii e d iii ii ii iii ce i i i n/a n/a n/a ii i i a a n/a 11 b n/a n/a n/a i ii i e d ii ii ii ii f i iii ii c ii c iii ii i b b i 12 b n/a n/a n/a iii ii i ae d ii i ii iii ac ii iii ii b n/a a ii ii iii b a n/a 13 d b c a i iii ii e d iii ii iii ii cef i i i n/a n/a n/a ii i i c c i 14 d a c a i iii ii ae d ii i ii iii ef ii i i n/a n/a n/a ii ii ii a a n/a 15 d a c a ii iii ii e e iii ii ii iii c i i ii c ii c iii ii i c c i

16 b n/a n/a n/a i ii ii e d ii ii ii iii c i ii i n/a n/a n/a iii ii ii d a n/a 17 d a c a i iii ii e e i i iii iii f i i i n/a n/a n/a iii i i a a n/a 18 a n/a n/a n/a n/a ii ii e a ii iii i i a iii i i n/a n/a n/a iii ii i d a n/a 19 a n/a n/a n/a n/a iii ii e d i ii ii i e ii iii i n/a n/a n/a iii i i c c i 20 d a c b i iii ii e e ii i ii iii f iii iii ii b n/a a iii ii i a a n/a 21 b n/a n/a n/a i ii ii c a ii iii i i ac iii ii i n/a n/a n/a i ii ii a a n/a 22 a n/a n/a n/a n/a ii iii e d ii ii ii iii af iii i i n/a n/a n/a i ii ii a a n/a Cantometrics Consensus Tape track number track Tape Consensus Cantometrics 23 d b c a ii iii ii e e iii ii ii ii def i i iii c ii abc ii ii i d a n/a 24 d a c a ii iii ii e e ii i i i a i ii ii c i ac iii i i d a n/a 25 b n/a n/a n/a i ii i e d iii ii i ii c i iii ii b n/a a iii ii i b b iii 26 b n/a n/a n/a ii iii ii e d i ii ii ii ac iii iii i n/a n/a n/a iii ii iii d a n/a 27 d a c b ii iii ii e a i i ii iii a i iii iii b n/a b iii i i d c iii 28 d c b n/a i ii ii e e ii i ii ii cf i ii i n/a n/a n/a iii ii i d c i 29 c n/a n/a n/a i iii iii e a ii i i i a iii iii iii c ii c iii i i d c ii 30 b n/a n/a n/a ii iii iii e e iii ii ii ii ae i i ii c ii c ii ii ii d b ii

23

CantoCore: A New Cross-Cultural Song Classification Scheme Appendix B: Inter-rater reliability Table B1. Inter-rater reliability values for song-structure characters from CantoCore and Cantometrics. See text for a description of κ as a measurement of reliability.

Character Line number Reliability (κ) CantoCore Cantometrics CantoCore Cantometrics Meter 1 11 0.36 0.043 Number of beats 2 n/a 0.60 n/a Beat sub-division 3 n/a 0.08 n/a Number of sub-beats 4 n/a undefined n/a Syncopation 5 n/a 0.35 n/a Motivic redundancy 6 n/a 0.22 n/a Durational variability 7 n/a 0.32 n/a Tonality 8 n/a undefined n/a Mode 9 n/a 0.25 n/a Number of pitch classes 10 n/a 0.38 n/a Hemitonicity 11 n/a 0.20 n/a Melodic interval size 12 21 0.48 0.36 Melodic range 13 20 0.40 0.33 Melodic contour 14 15 0.37 0.19 Melisma 15 29 0.81 0.42 Vocables 16 10 0.53 0.62 Number of vocal parts 17 4 0.68 0.15 Rhythmic texture 18 12 0.62 0.33 Harmonic texture 19 n/a 1.00 n/a Relative motion 20 22 0.25 0.14 Phrase repetition 21 16 0.70 0.23 Phrase length 22 17 0.39 0.22 Phrase symmetry 23 18 0.51 0.35 Solo/group arrangement 24 1 0.62 0.48 Responsorial arrangement 25 n/a 0.50 n/a Phrase overlap 26 n/a 0.64 n/a Position of the final tone n/a 19 n/a 0.36

24 Analytical Approaches To World Music 2.1 (2012) 87-137 Table B2. Inter-rater reliability values for performance-style characters (Cantometrics only). See text for a description of κ as a measurement of reliability.

Character Line number Reliability (κ) Tonal blend 5 0.49 Rhythmic blend 6 0.29 Embellishment 23 0.19 Tempo 24 0.14 Volume 25 0.66 Rubato 26 0.31 Glissando 28 0.13 Tremolo 30 0.46 Glottal shake 31 0.20 Register 32 0.25 Vocal width 33 0.28 Nasalization 34 0.15 Raspiness 35 0.12 Accent 36 0.24 Enunciation 37 0.42

Table B3. Inter-rater reliability values for instrumentation characters (Cantometrics only). See text for a description of κ as a measurement of reliability.

Character Line number Reliability (κ) Relationship to voice 2 0.14 Responsorial arrangement 3 0.22 Number of instrumental parts 7 0.07 Tonal blend 8 omitted (Lomax 1976) Rhythmic blend 9 -0.20 Meter 13 0.34 Rhythmic texture 14 0.16 Rubato 27 0.04

25